On 05.12.18 03:34, Wei Yang wrote:
> Currently locking for memory hotplug is a little complicated.
> 
> Generally speaking, we leverage the two global lock:
> 
>   * device_hotplug_lock
>   * mem_hotplug_lock
> 
> to serialise the process.
> 
> While for the long term, we are willing to have more fine-grained lock
> to provide higher scalability.
> 
> This patch divides Locking Internal section based on these two global
> locks to help readers to understand it. Also it adds some new finding to
> enrich it.
> 
> [David: words arrangement]
> 
> Signed-off-by: Wei Yang <richard.weiy...@gmail.com>
> ---
>  Documentation/core-api/memory-hotplug.rst | 27 ++++++++++++++++++++++++---
>  1 file changed, 24 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/core-api/memory-hotplug.rst 
> b/Documentation/core-api/memory-hotplug.rst
> index de7467e48067..95662b283328 100644
> --- a/Documentation/core-api/memory-hotplug.rst
> +++ b/Documentation/core-api/memory-hotplug.rst
> @@ -89,6 +89,20 @@ NOTIFY_STOP stops further processing of the notification 
> queue.
>  Locking Internals
>  =================
>  
> +There are three locks involved in memory-hotplug, two global lock and one 
> local
> +lock:
> +
> +- device_hotplug_lock
> +- mem_hotplug_lock
> +- device_lock
> +

Do we really only ever use these three and not anything else when
adding/removing/onlining/offlining memory?

(I am thinking e.g. about pgdat_resize_lock)

If so, you should phrase that maybe more generally Or add more details :)

"In addition to fine grained locks like pgdat_resize_lock, there are
three locks involved ..."


> +Currently, they are twisted together for all kinds of reasons. The following
> +part is divided into device_hotplug_lock and mem_hotplug_lock parts
> +respectively to describe those tricky situations.
> +
> +device_hotplug_lock
> +---------------------
> +
>  When adding/removing memory that uses memory block devices (i.e. ordinary 
> RAM),
>  the device_hotplug_lock should be held to:
>  
> @@ -111,13 +125,20 @@ As the device is visible to user space before taking 
> the device_lock(), this
>  can result in a lock inversion.
>  
>  onlining/offlining of memory should be done via device_online()/
> -device_offline() - to make sure it is properly synchronized to actions
> -via sysfs. Holding device_hotplug_lock is advised (to e.g. protect 
> online_type)
> +device_offline() - to make sure it is properly synchronized to actions via
> +sysfs. Even mem_hotplug_lock is used to protect the process, because of the
> +lock inversion described above, holding device_hotplug_lock is still advised
> +(to e.g. protect online_type)
> +
> +mem_hotplug_lock
> +---------------------
>  
>  When adding/removing/onlining/offlining memory or adding/removing
>  heterogeneous/device memory, we should always hold the mem_hotplug_lock in
>  write mode to serialise memory hotplug (e.g. access to global/zone
> -variables).
> +variables). Currently, we take advantage of this to serialise sparsemem's
> +mem_section handling in sparse_add_one_section() and
> +sparse_remove_one_section().
>  
>  In addition, mem_hotplug_lock (in contrast to device_hotplug_lock) in read
>  mode allows for a quite efficient get_online_mems/put_online_mems
> 


-- 

Thanks,

David / dhildenb

Reply via email to